Diﬀ erent strategies of sequential and combination chemotherapy for patients with poor prognosis advanced colorectal cancer (MRC FOCUS): a randomised controlled trial

(1)

Diﬀ erent strategies of sequential and combination

chemotherapy for patients with poor prognosis advanced colorectal cancer (MRC FOCUS): a randomised controlled trial

Matthew T Seymour, Timothy S Maughan, Jonathan A Ledermann, Clare Topham, Roger James, Stephen J Gwyther, David B Smith, Stephen Shepherd, Anthony Maraveyas, David R Ferry, Angela M Meade, Lindsay Thompson, Gareth O Griﬃ ths, Mahesh K B Parmar, and Richard J Stephens, for the FOCUS Trial Investigators* and the National Cancer Research Institute Colorectal Clinical Studies Group

Summary

Background In the non-curative setting, the sequence in which anticancer agents are used, singly or in combination, may be important if patients are to receive the maximum period of disease control with the minimum of adverse eﬀ ects. We compared sequential and combination chemotherapy strategies in patients with unpretreated advanced or metastatic colorectal cancer, who were regarded as not potentially curable irrespective of response.

Methods We studied patients with advanced colorectal cancer, starting treatment with non-curative intent.

2135 unpretreated patients were randomly assigned to three treatment strategies in the ratio 1:1:1. Strategy A (control group) was single-agent fl uorouracil (given with levofolinate over 48 h every 2 weeks) until failure, then single-agent irinotecan. Strategy B was fl uorouracil until failure, then combination chemotherapy. Strategy C was combination chemotherapy from the outset. Within strategies B and C, patients were randomly assigned to receive, as the combination regimen, fl uorouracil plus irinotecan (groups B-ir and C-ir) or fl uorouracil plus oxaliplatin (groups B-ox and C-ox). The primary endpoint was overall survival, analysed by intention to treat. This study is registered as an International Standard Randomised Controlled Trial, number ISRCTN 79877428.

Results Median survival of patients allocated to control strategy A was 13·9 months. Median survival of each of the other groups was longer (B-ir 15·0, B-ox 15·2, C-ir 16·7, and C-ox 15·4 months). However, log-rank comparison of each group against control showed that only C-ir—the ﬁ rst-line combination strategy including irinotecan—satisﬁ ed the statistical test for superiority (p=0·01). Overall comparison of strategy B with strategy C was within the predetermined non- inferiority boundary of HR=1·18 or less (HR=1·06, 90% CI 0·97–1·17).

Interpretation Our data challenge the assumption that, in this non-curative setting, maximum tolerable treatment must necessarily be used ﬁ rst-line. The staged approach of initial single-agent treatment upgraded to combination when required is not worse than ﬁ rst-line combination, and is an alternative option for discussion with patients.

Introduction

Advanced colorectal cancer causes over half a million deaths every year worldwide.¹ This death toll could fall with advances in prevention, screening, and curative treatment;

however, there are—and will remain—many patients with advanced disease for whom our realistic therapeutic goal is not complete cure, but rather to control the cancer suﬃ ciently to postpone death and reduce symptoms, with toxic eﬀ ects and inconvenience for the patient kept to a minimum.

Until 2005, colorectal cancer drug treatment largely centred on only three active agents: 5-fl uorouracil, oxaliplatin, and irinotecan. Several trials with fl uorouracil showed that its eff ectiveness is improved when given as a slow infusion,² with the addition of leucovorin,³ or using both of these methods.⁴ Irinotecan and oxaliplatin have mechanisms that diff er from fl uorouracil, and either can show non-crossresistance or synergy with this agent in some preclinical models.^5,6 Irinotecan was licensed in the UK in 1998 as single-agent treatment after fl uorouracil failure.⁷ By 2000, standard treatment for most patients treated within the UK National Health Service was

ﬁ rst-line ﬂ uorouracil followed, on progression, by single-agent irinotecan.

Four trials in 1999–2000 randomly assigned patients with metastatic colorectal cancer to initial treatment with fl uorouracil alone or in combination with either irinotecan or oxaliplatin.^8–11 In all these trials, combination treatment improved the objective response rate and time to progression. Second-line treatment was not prespecifi ed and diff ered between groups, making interpretation of overall survival diffi cult, but the two trials with irinotecan showed improved overall survival with fi rst-line combination chemotherapy^8,9 whereas those with oxaliplatin did not.^10,11 Taken together, the four trials provided good evidence that the use of at least one of these drugs in addition to fl uorouracil improves survival in patients with metastatic colorectal cancer. There remained some doubt, however, whether fi rst-line combination treatment was essential to obtain this benefi t—might equally good results be achieved with less toxic eff ects and cost with a staged treatment approach, starting with fl uorouracil and adding or changing to other drugs after failure?

Lancet 2007; 370: 143–52 See Comment page 105 See Articles page 135

*Collaborators listed at end of report

Cookridge Hospital, Leeds, UK (Prof M T Seymour MD);

Velindre Hospital, Cardiﬀ , UK (Prof T S Maughan MD); UCL Hospitals, London, UK (Prof J A Ledermann MD); St Luke’s Cancer Centre, Guildford, UK (C Topham MD); Mid-Kent Oncology Centre, Maidstone, UK (Prof R James MD); East Surrey Hospital, Redhill, UK (S J Gwyther MBBS);

Clatterbridge Centre for Oncology, Wirral, UK (D B Smith MD); Cheltenham General Hospital, Cheltenham, UK (S Shepherd MD); Princess Royal Hospital, Hull, UK (A Maraveyas PhD); New Cross Hospital, Wolverhampton, UK (Prof D R Ferry MD); and the Medical Research Council Clinical Trials Unit, London, UK (A M Meade PhD,

L Thompson MSc, G O Griﬃ ths MSc, Prof M K B Parmar DPhil, R J Stephens) Correspondence to:

Prof Matt Seymour, Gastrointestinal Cancer Research Unit, Cookridge Hospital, Leeds LS16 6QB, UK

matt.seymour@leedsth.nhs.uk

(2)

In subsequent years, evidence has emerged for use of all three cytotoxic drugs during the course of a patient’s illness,^12,13 and for adding targeted treatments against vascular endothelial growth factor ¹⁴ and epidermal growth factor receptor.¹⁵ However, the underlying question remains—in the context of non-curative cancer treatment,

how should our active drugs be sequenced to provide patients with the maximum duration of disease control and minimum of adverse eﬀ ects?

In May, 2000, the UK Medical Research Council launched FOCUS (Fluorouracil, Oxaliplatin, and CPT11 [irinotecan]—Use and Sequencing). We aimed to establish the best sequence of the ﬁ rst two cytotoxic drugs, ﬂ uorouracil and either irinotecan or oxaliplatin, when treating patients with poor prognosis advanced colorectal cancer.

Methods

Patients

Patients were required to have histologically conﬁ rmed colorectal adenocarcinoma, with inoperable metastatic or locoregional disease. The trial targeted patients with poor prognosis—namely, those with inoperable disease for whom the clinician judged that, even if there was a response to chemotherapy, curative surgery would still not be feasible. Eligibility criteria included: disease measurable by RECIST (response evaluation criteria in solid tumours);¹⁶ WHO performance status 0–2;¹⁷ no previous chemotherapy for metastatic disease; white blood count greater than 4×10⁹/L, platelet count greater than 150×10⁹/L, serum bilirubin concentration less than 1·25×upper limit of normal, alkaline phosphatise concentration less than 5×upper limit of normal, calculated glomerular ﬁ ltration rate or EDTA (edetic acid) clearance greater than 50 mL per min; and older than 18 years. Exclusion criteria included uncontrolled medical comorbidity likely to compromise treatment.

Study design and interventions

Patients were randomly assigned with a minimisation procedure to stratify for clinician, performance status, primary tumour resected or in situ, and distant metastases present or absent.

Three treatment strategies were compared, each with a diﬀ erent sequence of use for the ﬁ rst two cytotoxic drugs.

Strategy A was the control group. First-line treatment was with ﬂ uorouracil (using the FU regimen¹⁸ deﬁ ned in table 1) and continued until evidence of treatment failure.

Thereafter, in patients ﬁ t enough for second-line treatment, single-agent irinotecan was given (with the Ir regimen,⁷ as deﬁ ned in table 1). Strategy B was deferred combination

Fluorouracil Irinotecan Irinotecan/ﬂ uorouracil Oxaliplatin/ﬂ uorouracil

Regimen FU Ir IrFU OxFU

Intravenous drug schedule

Levofolinate 175 mg (2 h), then FU 400 mg/m² (bolus) FU 2800 mg/m² (46 hr)

Irinotecan 350 mg/m² (30–90 min). (300 mg/m² if age >70 years or performance status 2)

Irinotecan 180mg/m² (30 min), then levofolinate 175 mg (2 h), then FU 400 mg/m² (bolus) FU 2400 mg/m² (46 h)

Oxaliplatin 85mg/m² plus levofolinate 175 mg (concurrent, 2 h), then FU 400 mg/m² (bolus) FU 2400 mg/m² (46 h) Dexamethasone 8 mg intravenously bolus day 1.

Oral days 2–4, (decreasing course)

8 mg intravenously bolus day 1.

Cycle repeat 14 days 21 days 14 days 14 days

For full details of regimens used please see references as follows: FU,¹⁸ Ir,⁷ IrFU,¹⁹ and OxFU.¹⁸

Table 1: Chemotherapy regimens

Strategy A Strategy B Strategy C

Group A Group B-ir Group B-ox Group C-ir Group C-ox

Total patients 710 356 356 356 357

Male 494 (70%) 244 (69%) 235 (66%) 240 (67%) 247 (69%)

Age (years) 63 (56–69) 64 (57–70) 64 (56–69) 64 (57–69) 64 (56–69) Primary site*

Colon 487 (69%) 232 (65%) 240 (67%) 226 (63%) 234 (66%)

Rectum 222 (31%) 123 (35%) 114 (32%) 130 (37%) 123 (34%)

Prior adjuvant chemotherapy 163 (23%) 96 (27%) 89 (25%) 94 (26%) 94 (26%) WHO performance status

0 294 (41%) 147 (41%) 147 (41%) 147 (41%) 148 (41%)

1 355 (50%) 181 (51%) 178 (50%) 179 (50%) 179 (50%)

2 61 (9%) 28 (8%) 31 (9%) 30 (8%) 30 (8%)

Baseline white blood cell count

<10x10⁹/L 519 (73%) 271 (76%) 253 (71%) 260 (73%) 266 (75%)

≥10x10⁹/L 183 (26%) 82 (23%) 99 (28%) 93 (26%) 91 (25%)

Distant metastases

Synchronous 398 (56%) 198 (56%) 187 (53%) 203 (57%) 210 (59%)

Metachronous 297 (42%) 150 (42%) 161 (45%) 147 (41%) 138 (39%)

None (local disease) 15 (2%) 8 (2%) 8 (2%) 16 (4%) 9 (3%)

Number of disease sites

1 235 (33%) 126 (36%) 123 (35%) 111 (31%) 103 (29%)

2 295 (42%) 142 (40%) 143 (40%) 154 (43%) 169 (47%)

>2 180 (25%) 88 (25%) 90 (25%) 91 (26%) 85 (24%)

Disease sites

Liver 527 (76%) 266 (76%) 266 (76%) 279 (80%) 275 (79%)

Lymph nodes 296 (43%) 116 (33%) 130 (37%) 119 (34%) 118 (34%)

Lungs 237 (34%) 130 (37%) 118 (34%) 130 (37%) 135 (39%)

Peritoneum 94 (14%) 46 (13%) 51 (15%) 45 (13%) 52 (15%)

Primary unresected 176 (25%) 92 (26%) 90 (25%) 89 (25%) 87 (24%)

Local recurrence 27 ( 4%) 15 ( 4%) 14 (4%) 15 (4%) 16 (4%)

Other sites 91 (13%) 52 (15%) 49 (14%) 53 (15%) 50 (14%)

Measurable disease 688 (97%) 340 (96%) 346 (97%) 340 (96%) 349 (98%)

Unmeasurable disease 22 (3%) 16 (4%) 10 (3%) 16 (4%) 8 (2%)

Data are number (%) or median (IQR). *Data missing or uncertain for primary site in group A (one patient), B-ir (one patient), and B-ox (two patients).

Table 2: Baseline patient characteristics

(3)

chemotherapy. Again, fi rst-line treatment was with fl uorouracil, but patients fi t enough for second-line treatment were to receive combination chemotherapy.

Strategy B was further divided into two groups, B-ir and B-ox, in a 1:1 ratio at the time of initial randomisation. For patients in group B-ir, the planned second-line combination was irinotecan and fl uorouracil, whereas in group B-ox it was oxaliplatin and fl uorouracil (using the IrFU¹⁹ and OxFU¹⁸ regimens, respectively; table 1). Strategy C was fi rst-line combination treatment. These patients received

combination chemotherapy from the outset, continued until treatment failure. This strategy was also subdivided, in a 1:1 ratio, into two groups, with patients in group C-ir receiving irinotecan and ﬂ uorouracil, and those in group C-ox receiving oxaliplatin and ﬂ uorouracil.

Treatment was started as soon as possible after randomisation and continued until treatment failure.

Breaks (eg, for holidays) were not allowed during the ﬁ rst 3 months, and were restricted to 4 weeks during the second 3 months. Thereafter, on the basis of evidence

2135 patients randomly assigned

710 allocated to group A

356 allocated to group B-ir

356 allocated to group B-ox

356 allocated to group C-ir

357 allocated to group C-ox

351 received FU

201 received OxMdG

342 received IrFU

344 received OxFU 346 received

FU

185 received IrFU 696 received

FU

364 received Ir 666 failed

ﬁrst-line treatment

339 failed ﬁrst-line treatment

346 failed second-line treatment

153 received salvage chemotherapy

178 received salvage chemotherapy 322 failed

ﬁrst-line treatment

30 not yet failed 7 not yet failed 8 not yet failed 20 not yet failed 17 not yet failed

18 not yet failed 8 not yet failed 5 not yet failed

6 received alternative ﬁrst-line regimen 6 died/progressed to

terminal care 2 missing data

251 died/progressed to terminal care 51 received alternative

2nd line regimen

193 died without salvage chemotherapy

149 died without salvage chemotherapy 5 received alternative

ﬁrst-line regimen 2 died/progressed to

3 received alternative ﬁrst-line treatment 2 missing data

8 received alternative ﬁst-line regimen 5 died/progressed to

3 received alternative ﬁrst-line regimen 7 died/progressed to

First-line treatment

Second-line treatment

Figure 1: Trial proﬁ le

(4)

from our group’s previous trial,²⁰ patients with responding or stable disease were allowed to pause their treatment, resuming the same treatment thereafter provided that progression did not take place within 12 weeks of the last treatment. In the sequenced strategies (A and B), the planned second-line treatment was started, provided that the patient met the fi tness criteria of the regimen, at the fi rst evidence of progression during—or within 12 weeks if pausing—fi rst-line fl uorouracil.

Detailed dose reduction and delay protocols were provided, which, in general, required a 1 week delay for unresolved toxic eff ects of grade 2 or more, and a 20% dose reduction after toxic eff ects of grade 3 or more, or after two delays for grade 2 toxic eff ects. Patients receiving oxaliplatin plus fl uorouracil were carefully monitored for neurosensory toxicity, and oxaliplatin was suspended from the schedule if symptoms of grade 2 or more persisted between cycles.

When the full allocated treatment plan had failed, further options were considered, including purely symptomatic treatment or salvage chemotherapy. Before December, 2002, the recommended salvage in all groups was infusional fl uorouracil plus mitomycin-C,²¹ avoiding crossover. Thereafter, with evidence emerging to support sequential use of both oxaliplatin and irinotecan,^12,13 the protocol was modifi ed. The new recommendation was for patients who had received irinotecan in our study to be off ered oxaliplatin in combination with either fl uorouracil or capecitabine, and for those who had received oxaliplatin to have irinotecan with fl uorouracil or capecitabine, in both cases with fortnightly schedules.²²

Multicentre and institutional research ethics committee approval was obtained, and patients provided written informed consent. Trial management was by the MRC Clinical Trials Unit, with MRC good clinical research practice,²³ overseen by an independent trial steering

committee. An independent data monitoring committee met yearly with a remit including closure of the trial or individual groups if, in their opinion, a diﬀ erence had emerged that would aﬀ ect patient management.

Tumour assessment was done at baseline (≤4 weeks before treatment), then every 12 weeks, with CT or MRI, scored by RECIST.¹⁶ However, responses were not routinely conﬁ rmed by repeat scans, and external radiological review was not undertaken. Toxic eﬀ ects were scored with National Cancer Institute common terminology criteria (version 2.0).

Quality of life was assessed at baseline, 6 and 12 weeks, and every 12 weeks thereafter, with the European Organisation for Research and Treatment of Cancer core quality of life questionnaire (EORTC QLQ-C30).²⁴ Overall survival was calculated from randomisation to death by any cause. Progression-free survival was calculated from randomisation to ﬁ rst evidence of progression or death from any cause.

Statistical methods

For time-to-event outcomes, Kaplan-Meier curves were produced, patients alive without the event having occurred were censored at the time last seen, and log-rank tests were used to compare curves. Hazard ratios and 95% CIs were calculated for all comparisons. Exploratory subgroup analyses were done according to the baseline characteristics listed in table 2.

The planned sample size was 2100 patients: 700 in every treatment strategy (A, B, and C) and 350 in each subgroup of strategies B and C. An anticipated 2-year survival of 15% in the control group would detect an improvement of 7·5% (to 22·5%) in any pair-wise comparison of control versus an individual novel group (1050 patients, one-sided log-rank, 80% power, 1% signiﬁ cance to correct for multiple com parisons). However, for consistency with other studies, a two-sided log-rank test is presented for these comparisons.

Since starting FOCUS, standard practice has changed internationally and in the UK, altering the emphasis of the questions posed by the trial. First-line treatment with ﬂ uorouracil, although remaining standard in the UK until 2005, was replaced by combination therapy from mid-2000 in many countries (and subsequently in some countries by combination treatment with bevacizumab), making strategy C more relevant to standard practice.

Therefore, before completion of accrual, we decided to add a supplementary analysis to those speciﬁ ed in the original protocol, taking strategy C as the reference and examining strategy B for non-inferiority. This analysis is based on a CI approach, since the trial was not originally designed to answer a non-inferiority question. It assumes no interaction between strategy and second drug (oxaliplatin or irinotecan). A reduction in 2-year survival of 5% (eg, from 20% to 15%), assuming proportional hazards, equates to a hazard ratio (HR) of about 1·18.

With the CI approach with a one-sided signiﬁ cance level of 5% we formed a 90% CI for the HR; non-inferiority as

First-line treatment Second-line treatment

FU IrFU OxFU Ir IrFU OxFU

Study groups A, B-ir, B-ox C-ir C-ox A B-ir B-ox

Patients assessed 1305 337 339 349 180 199

Neutropenia 118 (9%) 65 (19%) 94 (28%) 43 (12%) 32 (18%) 50 (25%) Nausea or vomiting 55 (4%) 32 (10%) 31 (9%) 31 (9%) 9 (5%) 14 (7%)

Stomatitis 25 (2%) 6 (2%) 6 (2%) 3 (1%) 2 (1%) 6 (3%)

Diarrhoea 74 (6%) 38 (12%) 34 (10%) 58 (17%) 14 (8%) 16 (8%)

Hand/foot syndrome 22 (2%) 4 (1%) 4 (1%) 2 (1%) 2 (1%) 6 (1%)

Sensory neuropathy 11 (1%) 5 (2%) 34 (10%) 3 (1%) 2 (1%) 6 (3%)

Alopecia 3 (<1%) 8 (2%) 3 (1%) 34 (10%) 5 (3%) 0 (0%)

Lethargy 174 (13%) 66 (20%) 73 (21%) 59 (17%) 37 (21%) 41 (20%)

Pain 176 (14%) 73 (22%) 60 (18%) 79 (23%) 26 (14%) 39 (20%)

Treatment-related death 11 (1%) 3 (1%) 4 (1%) 3 (1%) 1 (1%) 2 (1%) 60-day all-cause mortality 52 (4%) 17 (5%) 14 (4%) 31 (9%) 13 (7%) 12 (6%)

Data are the number (%) of patients who received at least one dose of each regimen who experienced an adverse event of grade 3 or worse, scored using National Cancer Institute common terminology criteria (version 2.0). For details of drug regimens please see table 1. For details of study groups please see text and ﬁ gure 1.

Table 3: Grade 3 or 4 adverse events reported

(5)

deﬁ ned above needs exclusion of an upper conﬁ dence limit 1·18 or more.

This study is registered as an International Standard Randomised Controlled Trial, number ISRCTN 79877428.

Role of the funding source

The study design, data collection, data analysis, data interpretation, writing of the report, and the decision to submit for publication involved employees of MRC Clinical Trials Unit. The corresponding author had full access to all the data in the study and had ﬁ nal responsibility for the decision to submit for publication.

Results

From May 1, 2000 to Dec 31, 2003, 2135 patients were randomly assigned at 59 centres in the UK and one in Cyprus. Table 2 shows baseline characteristics. Figure 1 shows the trial profi le. Patients in strategies A and B received a median of 11 cycles (range 1–51) of the allocated fi rst-line fl uorouracil regimen; patients assigned to irinotecan plus fl uorouracil in strategy C received a median of 12 cycles (1–36). Patients in strategy C allocated to oxaliplatin plus fl uorouracil received a median of 12 cycles (1–58), of which oxaliplatin was included for 3506 of 3740 (94%) cycles (median 11 [1–58] per patient); the remainder were given as fl uorouracil alone after development of persistent neuropathy.

Figure 1 shows the numbers of patients in the two-stage strategies who proceeded to second-line chemotherapy. Of the 1348 in whom fl uorouracil had failed at the time of analysis, 750 (56%) had received the planned second-line regimen, and a further 131 (10%) received an alternative second-line regimen; however, 467 (35%) patients had moved to terminal care or died without receiving further chemotherapy after fi rst-line fl uorouracil. The median amount of time spent on second-line combination therapy was similar in all groups. Patients in group A received a median of four cycles (1–24) of irinotecan every 3 weeks.

Those in group B receiving irinotecan received a median of six cycles (1–23) of irinotecan plus ﬂ uorouracil every 2 weeks. Those in group B allocated second-line oxaliplatin plus ﬂ uorouracil, received a median of six cycles (1–24), of which oxaliplatin was included for 1582/1688 (94%) cycles (median six [1–21] per patient).

Figure 1 also shows the usage of salvage chemotherapy.

To date, 49% of patients (669/1368) in whom the allocated treatment has failed have received salvage chemotherapy.

The proportion was higher for those in strategy C (358/649 [55%]), for whom FOCUS treatment was a single line of treatment, than it was for those in strategy A and B (311/719 [43%]), who had already received two lines of therapy at the time of FOCUS failure (ﬁ gure 1). After the change in salvage therapy recommendations in December, 2002, FOCUS patients could receive all three

0 0·2 0·4 0·6 0·8 1·0

0 6 12 18 24 30 36

0 6 12

710 562 397 247 139 73 40 710 562 397 247 139 73 40 710 562 397 247 139 73 40

712 594 425 270 153 72 38 356 300 207 131 72 40 21 356 294 218 139 81 32 17

713 Strategy A Strategy B Strategy C

Group A

A B C

Group B-ir Group C-ir

Group A Group B-ox Group C-ox

598 448 294 174 84 42 356 291 223 148 93 49 23 357 307 225 146 81 35 19

Patients at risk

Survival

Months

18 24

Events Total 617 616 606

710 712 713

30 36

Group A Group B-ir Group C-ir

Events Total 617 300 295

710 356 356

Group A Group B-ox Group C-ox

Events Total 617 316 311

710 356 357

0 6 12 18 24 30 36

Figure 2: Kaplan-Meier plots for overall survival by intention to treat

(A) Survival by treatment strategy. (B) Survival by treatment group (control vs B-ir vs C-ir). (C) Survival by treatment group (control vs B-ox vs C-ox). See table 4 for log-rank statistics.

Log-rank test comparison

HR (95% CI)

p-value (two-sided test)

Median survival in reference group*

Diﬀ erence†

(95% CI) between reference group and comparitor Are any of the novel plans better than the control?

A vs B 0·94 (0·84 – 1·05) 0·24 13·9 0·9 (–0·7 to 2·6)

A vs B-ir 0·91 (0·79–1·03) 0·16 13·9 1·4 (–0·4 to 3·7)

A vs B-ox 0·97 (0·85–1·11) 0·65 13·9 0·4 (–1·4 to 2·5)

A vs C 0·88 (0·79 – 0·98) 0·02 13·9 1·9 (+0·3 to 3·7)

A vs C-ir 0·84 (0·73–0·96) 0·01 13·9 2·6 (+0·6 to 5·1)

A vs C-ox 0·93 (0·81–1·06) 0·26 13·9 1·1 (–0·8 to 3·3)

Does the choice of irinotecan or oxaliplatin aﬀ ect survival?

[B-ir + C-ir] vs [B-ox + C-ox]

1·09 (0·97–1·21) 0·14 15·8 –1·3 (–2·7 to 0·5)

B-ir vs B-ox 1·06 (0·91–1·24) 0·46 15·0 –0·8 (–2·9 to 1·5)

C-ir vs C-ox 1·12 (0·95–1·31) 0·18 16·7 –1·8 (–4·0 to 0·9)

*Group A is the reference group for whether any of the novel plans are better than the control. Irinotecan group is the reference group for whether the choice of irinotecan or oxaliplatin aﬀ ect survival. †Diﬀ erence calculated by application of log-rank HR to median survival in control group.

Table 4: Overall survival log-rank comparisons

(6)

drugs (ﬂ uorouracil, irinotecan, and oxaliplatin) at some point in their management. However, at the time of analysis, only 482 (23%) patients had done so. The proportion who had received all three drugs was higher for patients allocated to strategy C (233/713 [33%]) than to A (113/710 [16%]) or B (136/712 [19%]). The proportion was much the same in patients allocated to irinotecan groups in strategies B and C (175/712 [25%]) or oxaliplatin groups in strategies B and C (194/713 [27%]).

All regimens were fairly well tolerated and safe. Treat- ment delays or modifi cations were made in less than 40% of patients at any point for all treatment regimens with one exception; fi rst-line OxFU regimen was delayed or modifi ed in 50% of patients, usually for neurosensory or haematological toxic eff ects after several cycles. Table 3 shows grade 3 or 4 toxic eff ects. Neutropenia and neuropathy were the main toxicities of OxFU, although grade 3 neuropathy was uncommon in the second-line setting, in which treatment duration was shorter. Diarrhoea and alopecia were most common with single-agent irinotecan.

Diar rhoea was no more frequent with IrFU regimen than with OxFU. Pain and lethargy were common with all regimens, which might have been in part due to the underlying disease. However, their incidence was noticeably lower with the FU regimen than with the other regimens.

610 patients (29%) had a serious adverse event caused or probably caused by the trial drugs; a further 262 patients (12%) had serious adverse events related to venous access. 24 deaths (1·1% of patients) were reported as deﬁ nitely or probably caused or precipitated by FOCUS treatment, with no great diﬀ erence between the regimens.

18 of these deaths were in the 2093 patients receiving fi rst-line treatment (0·9% of patients), and six were in the 755 receiving second-line therapy (0·8%). Death occurred within 30 days of the fi nal dose of the fi rst-line chemotherapy in a further 130 (6·2%) patients (124 because of cancer; six from an unrelated medical illness), and within 30 days of the last second-line chemotherapy in a further 42 (5·6%) patients (39 because of cancer, three from intercurrent illness). There was no imbalance in all-cause mortality at day 60 (table 3).

Up until now, 1839 (86%) patients have died, and the median follow-up of the survivors is 26·5 months. Figure 2 shows the Kaplan-Meier overall survival curves by intention to treat for all patients randomly assigned. Strategies B and C produced very similar survival, both slightly better than strategy A. 2-year survival was 22% with strategy A, 25%

with strategy B, and 28% with strategy C. In pairwise log- rank tests (table 4), overall comparison of strategy C with control reached p=0·02, but did not satisfy the stringent level of p<0·01 required to conﬁ rm superiority in the context of multiple testing.

Figure 2 also shows overall survival by the individual treatment groups. All four groups in strategies B and C provided longer overall survival than did the control group, but only irinotecan used in the fi rst-line combination treatment was signifi cantly better (table 4). Comparison of irinotecan versus oxaliplatin, whether used in the fi rst-line combination setting, second-line combination, or at any time showed no signifi cant diff erence (table 4).

In response to changes in standard practice after the start of the trial, an additional non-inferiority analysis was

Log-rank test comparison

HR (90% CI)

Median survival (months)

Conﬁ dently excludes detriment with strategy B larger than:*

Strategy C (reference)

Strategy B

C vs B 1·06 (0·97–1·17) 15·9 15·1 2·3 months

C-ir vs B-ir 1·08 (0·94–1·24) 16·7 15·0 3·2 months C-ox vs B-ox 1·04 (0·92–1·19) 15·4 15·2 2·5 months

*Estimation of the largest detriment to the comparator group that cannot be reliably excluded. It is calculated by using the upper end of the 90% CI in the following way: comparator median–([1/upper end of 90%CI]×comparator median).

Table 5: Is deferred combination (strategy B) non-inferior to ﬁ rst-line combination (strategy C)?

First-line treatment Second-line treatment

Fluorouracil Irinotecan+

ﬂ uorouracil

oxaliplatin+

ﬂ uorouracil

Irinotecan Irinotecan+

ﬂ uorouracil

oxaliplatin+

ﬂ uorouracil

Study groups A, B-ir, B-ox C-ir C-ox A B-ir B-ox

Total number treated (receiving ≥1 dose) 1393 342 344 364 185 201

CR 57 19 29 8 1 3

PR 335 147 166 31 29 43

SD (≥12 weeks) 487 89 72 107 68 74

PD 249 40 40 149 52 50

Not assessed† 265 47 37 69 35 31

Response rate (CR+PR)‡ 28% 49% (p<0·001)§ 57% (p<0·001)§ 11% 16% (p=0·07)¶ 23% (p<0·001)¶

Disease control ≥12 weeks (CR+PR+SD)‡ 63% 75% (p<0·001)§ 78% (p<0·001)§ 40% 53% (p=0·004)¶ 60% (p<0·001)¶

Median PFS (months) 6·3 8·5 (p<0·001)§ 8·7 (p<0·001)§ 4·3 4·4 (p=0·75)¶ 4·8 (p=0·74)¶

RECIST=response evaluation criteria in solid tumours. PFS=progression free survival. CR=complete response. PR=partial response. SD=stable disease. PD=progressive disease.

*Responses did not need to be conﬁ rmed by a second scan. †Includes any reason for failure to reassess radiologically—eg, clinically obvious deterioration. ‡Denominator includes all patients who received one or more dose, whether or not subsequently assessed. §Compared with ﬂ uorouracil (χ² test for response rate and disease control;

log-rank test for PFS.) ¶Compared with irinotecan (χ² test for response rate and disease control; log-rank test for PFS).

Table 6: RECIST* response and progression-free survival

(7)

added, to compare deferred combination treatment (strategy B) with the new international standard of ﬁ rst- line combination (strategy C). When analysed for non- inferiority with reference to strategy C, strategy B gave a HR of 1·06 (90% CI 0·97–1·17). These data exclude an inferiority margin of HR 1·18 or more, corresponding to a reduction of more than 5% in 2-year survival or a diﬀ erence in median survival of more than 2·3 months. The results for the two individual drugs irinotecan and oxaliplatin are similar (table 4), but with fewer patients the CIs are wider than for the grouped comparison, so these individual comparisons do not have adequate power to conclude non-inferiority (table 5).

Table 6 shows RECIST responses and median progression-free survival for all treatment phases, and

should be interpreted in the context that scans were not reviewed externally. As anticipated from previous trials, the response rates and progression-free survival for fi rst-line IrFU and OxFU regimens were signifi cantly better than for fl uorouracil alone. For patients in strategy A or B who went on to receive their allocated second-line treatment, the combination therapies IrFU and OxFU both gave higher response rates than did irinotecan alone, signifi cantly so in the case of OxFU, although the rates of progression-free survival were not signifi cantly improved.

During the fi rst 18 months from randomisation, the mean WHO performance status fell from 0·7 to about 1·1 but no diff erences were seen between the fi ve treatment groups. Detailed quality of life data were obtained and will be reported elsewhere. However, the mean overall quality

Strategy B better Strategy C better

0 1 2 0·5 1·5

Age (years)

Interaction p=0·51 Sex

Interaction p=0·21

WHO performance status

Interaction p=0·38 Primary site

Interaction p=0·31 Prior adjuvant ﬂuorouracil

Interaction p=0·40

WBC

Interaction p=0·10 Number of disease sites

Interaction p=0·97 Type of disease

Interaction p=0·18 Number events/number entered

Strategy B Strategy C

O–E Variance HR (95% CI)

Male 405/480 423/488 3·11 206·14 1·02 (0·89–1·16) p=0·828 Female 211/231 183/225 16·54 97·89 1·18 (0·97–1·44) p=0·095

60–69 258/304 257/290 –9·48 127·99 0·93 (0·78–1·10) p=0·402

Colon 412/467 392/452 3·27 200·09 1·02 (0·88–1·17) p=0·817

<60 203/230 201/252 19·07 99·65 1·21 (1·00–1·47) p=0·056

Rectum 200/237 206/253 14·21 100·00 1·15 (0·95–1·40) p=0·155

Yes 155/185 159/189 –1·61 77·97 0·98 (0·78–1·22) p=0·855 No 460/523 447/523 20·51 225·33 1·10 (0·96–1·25) p=0·172

0 243/293 240/295 7·34 120·30 1·06 (0·89–1·27) p=0·503

1 318/358 310/358 5·93 156·18 1·04 (0·89–1·22) p=0·635 2 55/59 56/60 9·52 26·13 1·44 (0·98–2·11) p=0·063

<10×10⁹/L 444/526 438/526 4·48 219·78 1·02 (0·89–1·16) p=0·763

≥ 10×10⁹/L 169/181 166/184 19·16 80·84 1·27 (1·02–1·58) p=0·033

1 195/242 157/205 5·84 86·91 1·07 (0·87–1·32) p=0·531

>2 169/185 168/186 6·03 83·69 1·07 (0·87–1·33) p=0·510

Measurable 594/684 583/687 25·26 292·98 1·09 (0·97–1·22) p=0·140 Unmeasurable 21/25 20/22 –3·46 9·81 0·70 (0·38–1·31) p=0·270

Total 616/709 606/711 19·68 304·32 1·07 (0·95–1·19) p=0·259

2 252/285 281/322 18·66 130·36 1·15 (0·97–1·37) p=0·102 70+ 155/178 148/171 8·64 74·94 1·12 (0·89–1·41) p=0·318

Figure 3: Forest plot of overall survival with strategy B or strategy C (according to subgroups) Horizontal tails indicate 99% and 95% CIs. O-E=observed minus expected results. WBC=white blood cell count.

(8)

of life score (combining questions 29 and 30 from the EORTC QLQ-C30) varied very little over time or between regimens; in particular we were not able to detect any advantage or disadvantage at 3 and 6 months associated with ﬁ rst-line combination chemotherapy (strategy C).

There was no defi nite evidence that the eff ect of treatment on survival was diff erent in any of the subgroups of patients defi ned by the baseline characteristics. Figure 3 shows the HRs for strategy B versus strategy C according to the baseline characteristics in table 2. No clear evidence of a diff erential benefi t of strategy C over strategy B was noted in these categories; however, a non-signifi cant trend was seen favouring strategy C over B in patients with WHO performance status 2.

Discussion

This large randomised trial has produced a surprising result which challenges accepted standard treatment approaches in advanced colorectal cancer therapy. It shows that, in the context of non-curative treatment, the strategy of initial ﬂ uorouracil alone, reserving combination therapy for second-line treatment, does not compromise patients’

survival or quality of life compared with fi rst-line com bin- ation chemotherapy. Several previous trials have compared fi rst-line fl uorouracil or combination therapy, but our trial prospectively compares the same drugs used in diff erent sequence strategies. Two subsequent trials have used similar designs. CAIRO-1 (Dutch Colorectal Cancer Group),²⁵ reported in this issue, compares sequenced single-agent capecitabine then irinotecan or fi rst-line irinotecan plus capecitabine (similar to groups A and C-ir in this trial), in both cases with planned oxaliplatin-based salvage therapy. FFCD 2000-05 (Federation Francophone de la Cancerologie) also compares sequence strategies and may add further supportive data to our fi ndings.

Improved treatment of metastatic colorectal cancer over the past decade has produced two linked but distinct beneﬁ ts for patients. First, a small but increasing number of patients can now be treated with a possibility of complete cure, usually through a combination of surgery and drug treatment.²⁶ Second, for patients who are not cured, new therapies oﬀ er the probability of increased survival, including a longer period with good quality of life.

These distinct goals—curative and non-curative—might be served by the same drug treatment strategy, but this is not necessarily so. Curative intent demands the maximum possible short-term anticancer eff ect; active agents need not be kept in reserve, and high toxic eff ects could be a worthwhile price for a major potential gain. By contrast, with non-curative treatment, longer-term scheduling to make best use of active agents throughout the illness is highly relevant and, with more modest potential gains, the negative eff ects of treatment are more important.

We addressed the clinical setting of non-curative treatment. From the outset, patients with operable metastases were excluded and clinicians were also encouraged not to enter patients into the study who might

become operable should they respond well. Subsequently, guidance from the UK National Institute of Clinical Excellence in 2002 mandated NHS-funded combination chemotherapy for any patient with liver metastases “that may become resectable following treatment”.²⁷ Consequently, patients accruing to FOCUS were increasingly selected as a poor-prognosis group who were deﬁ nitely incurable. This approach might explain why, despite excellent response and progression free survival data, the overall survival is lower in FOCUS than in contemporaneous trials that included all patients with metastatic disease.^28,29 Our ﬁ ndings should therefore be interpreted with careful reference to the trial’s population.

Clinicians and patients considering the staged treatment approach may be attracted by the prospect of a low risk of toxic eff ects during fi rst-line therapy; indeed, the fl uorouracil regimen that we used produced an extremely favourable profi le (table 3). However, staged therapy also brings two concerns. At the time of progressing on fi rst- line treatment, will the patient be unfi t for the planned second-line schedule, compromising their survival? And will the lower initial response rate—and a disease progression episode before introduction of the second drug—compromise their quality of life? In FOCUS, a third of patients allocated to staged strategies never received second-line therapy. This proportion could possibly have been reduced by more frequent CT scan surveillance.

However, despite this factor, which might have disadvantaged patients in staged treatment plans, we found that overall survival with strategy B was no worse than strategy C. Similarly, we detected no diﬀ erences in quality of life scores or performance status between the ﬁ rst-line combination and staged treatment approaches.

These fi ndings are hard to explain. Perhaps patients whose condition deteriorates most rapidly on fl uorouracil alone are also least likely to benefi t from combination therapy.

Or perhaps the staged treatment approach actually improves the survival of patients with fl uorouracil-sensitive disease, off -setting and so masking a detriment in patients who are fl uorouracil-resistant.

A molecular marker sub-study of FOCUS has reported that low tumour protein expression of topoisomerase-1 is associated with low expression of thymidylate synthase, and that patients with this molecular profi le have good outcomes when given fl uorouracil alone but gain little additional benefi t from irinotecan or oxaliplatin.

Conversely, patients with high expression of these proteins have poor outcomes with fl uorouracil alone but major benefi t from receiving irinotecan or oxaliplatin fi rst-line.³⁰ These data need to be validated in other trial populations, but hold promise for molecular selection of the best therapeutic strategy for individual patients.

Compared with the control strategy of sequential single- agent ﬂ uorouracil then irinotecan, median survival times were longer with each of the novel groups. However, most of the diﬀ erences were minor, and in only one group—

ﬁ rst-line irinotecan plus ﬂ uorouracil—was the statistical

(9)

test for superiority signifi cant. This fi nding is consistent with the survival benefi t seen in previous trials of fl uorouracil versus fl uorouracil plus irinotecan in which second-line treatment, although not controlled, consisted of single-agent irinotecan for many patients assigned to the control group.^8,9 The results for irinotecan and oxaliplatin were not identical. The response rate with oxaliplatin plus fl uorouracil—both fi rst and second line—

was higher than it was with irinotecan plus ﬂ uorouracil;

however, overall survival tended to favour the strategies that included the irinotecan combination. There was no imbalance of third-drug salvage treatment to account for this observation, and early omission of oxaliplatin for neurotoxicity seems not to have been a major factor, since it was omitted from only 6% of cycles. Our fi ndings are not inconsistent with previous trials in which, despite its high response rate, fi rst-line oxaliplatin combinations have not shown defi nite survival benefi ts.^10,11,31

What is the relevance of FOCUS in 2007, after the arrival of targeted treatments against vascular endothelial growth factor and epidermal growth factor receptor. Oncologists in many countries are now not asking whether they should give one drug or two, but rather whether they should use three, four, or more in combination. The march of progress in oncology has generally assumed that the greatest anticancer activity today will automatically lead to the greatest benefi t tomorrow; keeping active treatment in reserve is for the fainthearted. This assumption is rarely challenged—few trials have methodically compared strategies of using drugs immediately or reserving them for later, and drug companies understandably prefer to support trials which bring their drug earlier in the treatment pathway. The result has been a progressive escalation of the complexity, toxicity, and cost of fi rst-line treatment. For patients with limited metastatic disease, this approach has certainly given an improved possibility of cure. But, for the larger number of patients with more extensive metastases who will not be cured, FOCUS off ers an important choice, informed by the knowledge that a decision to opt for a staged treatment approach, starting with less toxic therapy and keeping active agents in reserve, entails minimal, if any, compromise in survival. The addition of bevacizumab to single-agent fl uoropyrimidine treatment is safe and eff ective in fi rst-line therapy,³² but is usually reserved for patients thought to be unfi t for combination treatment; the FOCUS data support extending this approach to a wider range of patients. Furthermore, our study shows that it is ethical and reasonable, in this defi ned patient population, to design fi rst-line studies in which novel agents are added to optimum fl uoropyrimidine treatment instead of combination chemotherapy, thus reducing the burden of toxic eff ects for trial participants.

Contributors

MTS was chief investigator of the trial and chaired the trial management group. MTS, TSM, MKBP, and RJS designed the trial, wrote the protocol, and prepared the manuscript with input from all members of the trial management group. TSM, JAL, CT, and RJ were members of the trial management group and contributed patients to the trial. SJG was a

member of the trial management group and led in radiology aspects of the trial. DBS, SS, AM, and DRF contributed patients to the trial and commented on the manuscript. MKBP, AMM, LT, GOG, and RJS were responsible for the data management, analysis, statistical analysis. and interpretation.

Trial management group

MT Seymour (chair), TS Maughan, JA Ledermann, C Topham, R James, SJ Gwyther, C Button, D Blake, J Smith, M Sculpher, G Griﬃ ths, A Meade, RJ Stephens, L Thompson.

MRC clinical trials unit

S Beall, C Chung, S Clawson, T Cullum, S Hassan, C Johnson, B May, C Murphy, J Nuttall, J Pickering, L van Dyck.

Data monitoring committee

J Northover (chair), J Brown, JL Mansi, M Aapro, R Stout.

Trial steering committee

M Mason (chair), R Rudd, PWM Johnson.

Clinical investigators (institution [number of patients contributed]) D B Smith, P Clark, E Marshall, S Myint, S M O’Reilly, M Iqbal (Clatterbridge Centre for Oncology, UK [180]); T S Maughan, A Brewster, T D L Crosby, D Mort, N Iqbal (Velindre Hospital Cardiﬀ , UK [138]);

S Shepherd, D Farrugia, K Benstead, S Elyan, R Owen, R Counsell (Cheltenham General Hospital, UK [108]); C Topham, G Middleton (St Luke’s Cancer Centre, Guildford, UK [104]); M Seymour, A Anthoney, D Sebag-Monteﬁ ore (Cookridge Hospital, Leeds, UK [89]); M Saunders, E Levine, J Valle, H Anderson, M Wilson, R E Hawkins (Christie Hospital, Manchester, UK [88]); J Summers, M Hill, R James, S Beesley,

D G L Pickering, (Mid-Kent Oncology Centre, Maidstone, UK [83]);

D J Radstone, B Orr, P Kirkbride, M Marples, D Levy, K Dunn (Weston Park Hospital, Sheﬃ eld, UK [71]); R H Phillips, C Lowdell, P Riddle (Charing Cross Hospital, London, UK [66]); H Wasan, C Vernon, V Khoo (Hammersmith Hospital, London, UK [56]); A Maraveyas, M Lind (Princess Royal Hospital, Hull, UK [55]); J C van der Voet, N Wadd, N Storey (James Cook University Hospital, Middlesbrough, UK [52]);

J Stewart, C Macmillan (Northamptonshire Centre for Oncology, UK [51]);

M J Churn, R Mehra, R Allerton, D R Ferry (New Cross Hospital, Wolverhampton, UK [51]); S Falk, K Hopkins (Bristol Oncology Centre, UK [47]); D R Ferry (Russells Hall Hospital, UK [46]); A Robinson, P Leonard (Southend Hospital, UK [45]); N J Hodson, M Wilkins, A Webb, G Newman, G P Deutsch (Royal Sussex County Hospital, Brighton, UK [45]); A O’Callaghan, G Khoury (Portsmouth Oncology Centre, UK [40]);

J Bridgewater, S Karp (North Middlesex Hospital, UK [38]); S Gollins, N Stuart (Glan Clwyd Hospital, UK [38]); M Napier, C G Rowland, D Sheehan, P Bliss, A Hong (Royal Devon and Exeter Hospital, Exeter, UK [35]); P Mack (Diana Princess of Wales Hospital, Grimsby, UK [35]);

J Ledermann (UCL Hospitals, London, UK [34]); J K Joﬀ e (Huddersﬁ eld Hospital, UK [33); F N Daniel (Derriford Hospital, Plymouth, UK [33]);

P Chakraborti (Derbyshire Royal Infi rmary, UK [33]); S M Crawford (Airedale General Hospital, Keighley, UK [32]); C Wilson, P Corrie (Addenbrooke’s Hospital, Cambridge, UK [32]); C Bradley (Bradford Royal Infi rmary, UK [30]); D Papamichael (Bank of Cyprus Oncology Centre, Cyprus [30]); F A Adab (North Staff ordshire Royal Infi rmary, UK [24]);

A Weaver, A Jones (Churchill Hospital, Oxford, UK [23]); K McAdam, K Fife (Peterborough District Hospital, UK [22]); I Geh (Birmingham Heartlands Hospital, UK [22]); B Sizer (Essex County Hospital [20]);

N Bailey (Torbay Hospital, Torquay, UK [19]); T J Iveson (Royal South Hants Hospital, Southampton, UK [18]); M Soukop, G Wilson, M Gray, D Dunlop (Glasgow Royal Inﬁ rmary, UK [18]); F Lofts (St George’s Hospital, London, UK [13]); H Yosef (Hairmyres Hospital, East Kilbride, UK [13]); R Begent, T Meyer, D Hochauser (Royal Free Hospital, London, UK [11]); S Beesley (Conquest Hospital, St Leonards on Sea, UK [11]);

M J Churn (Kidderminster Hospital, UK [9]); J Glaholm (Good Hope Hospital, Sutton Coldﬁ eld, UK [9]); A T Axford (Bronglais General Hospital, Aberystwyth, UK [9]); N Stuart (Ysbyty Gwynedd, Bangor, UK [8]);

S Falk (Yeovil Hospital, UK [8]); J Ledermann (Whittington Hospital, London, UK [8]); D Landau, B Bryant (Queen Elizabeth Hospital, Woolwich, UK [8]); J M Bozzino (South Tyneside District Hospital, South Shields, UK [7]); S Tahir (Broomﬁ eld Hospital, Chelmsford, UK [6]);

T J Iveson (Salisbury District Hospital, UK [5]); F A Adab (Staﬀ ordshire General Hospital, UK [4]); C S Askill (Singleton Hospital, Swansea, UK